著者

湯浅 友幸 白山 晋

出版者

情報処理学会

雑誌

研究報告数理モデル化と問題解決(MPS) (ISSN:18840930)

巻号頁・発行日

vol.2010, no.29, pp.1-7, 2010-12-09

---

ネットワーク上の現象は大局的構造から説明されることが多く,局所的な見地から分析されることは少ない.本稿ではノードの性質に注目した新たな分析手法を提案する.提案手法では自己組織化マップ (SOM) を用いたノード分類に基づくシミュレーションの可視化と分析を行う.そして,2 つのシミュレーションモデルを用いた実験により提案手法の有効性を示す.Most researches concerning the influence of network strucuture on phenomena on the network are carried out based on relationships between global statistics of the network structure and characteristic properties of those phenomena, even though the local structure has a significant effect on dynamics of some pheomenon. In this paper, we propose a new analysis method for some phenomena on networks based on the categorization of nodes. First, local statistics such as the average path length, the clustering coefficient for a node are calculated and are given to each node. Secondly, the nodes are categorized by Self-Organizing Map (SOM) and are grouped. Characteristic properties of some phenomenon are visualized into the grouped nodes. From some numerical results using two simulation models, the validity of our method is examined.